AMMI methodology in soybean: Cluster analysis with bootstrap resampling in genetic divergence and stability

Resumo

This study aimed to propose a clustering methodology with bootstrap resampling using the Additive Main Effects
and Multiplicative Interaction Analysis (AMMI) to contribute to better prediction of phenotypic stability of genotypes
and environments. It also aims to analyze the genetic divergence in the assessment of soybean lines, identify genotypes
with high-yielding characteristics, with control of chewing and sucking insect pests, and cluster similar genotypes for
the traits evaluated. A total of 24 experiments were conducted in randomized blocks, with two replications subdivided in
experimental groups with common controls. AMMI with principal component analysis indicated that PC1 and PC2 were
significant, explaining 83.9% of the sum of squares of the interaction. The first singular axis of AMMI analysis captured
the highest percentage of “pattern” and, with subsequent accumulation of the dimensions of the axes, there was a
decrease in the percentage of “pattern” and an increase in “noise”. The Euclidean distance between genotype scores
was used as the dissimilarity measure and clusters were obtained by the hierarchical method of Ward. Genotypes 97-
8011, 97-8029, 97-8050 and IAS-5 had the best performance and are promising for recommendation purposes, with the
greatest stability and best performance on grain yield.